External Prometheus at 10.0.15.31 already runs the kubernetes-pods job
in prometheus.yml.j2; pods need prometheus.io/scrape + port + path +
job annotations to be picked up. The prop-metrics NodePort selector
also tightened to tier=hot so backfill (PHX_SERVER=false, no :5000
listener) is excluded from the scrape pool.
A dedicated Prometheus at 10.0.15.25 on the private network can now
scrape the cluster via `<any-node-ip>:30090/metrics`. Scrape rotates
across pods; BEAM metrics stay representative as a sample and the
oban queue-depth gauge reads from the shared DB so all pods report
the same value.
